cloudglue

The Cloudglue MCP Server enables large language models (LLMs) to interact with your Cloudglue video data. By exposing a set of tools, it lets you explore video collections, locate specific videos, and perform analysis such as transcripts, entity extraction, and chapter segmentation. The server is configured to run via the Node.js-based MCP runtime using npx to fetch the latest cloudglue-mcp-server package and start it with your API key. Tools exposed through the server include discovery and navigation functions like list_collections and list_videos, and individual video analysis tools such as describe_video and extract_video_entities. You can then connect MCP clients (for example Cursor or Claude Desktop) and execute commands to retrieve descriptions, transcripts, entities, and structured data from videos stored in Cloudglue.

Prerequisites:

• Node.js and npm installed on your machine
• A Cloudglue API Key

Installation steps:

1. Ensure Node.js and npm are installed. Verify with: node -v npm -v

2. Install dependencies and set up the server by pulling the latest package via npx as shown in the usage configuration, or clone the repository if provided (not required for MCP runtime).

3. Run the server via MCP configuration (see mcp_config). Example steps:

   • Save the following configuration to a file or pass it to your MCP client: { "mcpServers": { "cloudglue": { "command": "npx", "args": [ "-y", "@cloudglue/cloudglue-mcp-server@latest", "--api-key", "<CLOUDGLUE-YOUR-API-KEY>" ] } } }

4. Replace <CLOUDGLUE-YOUR-API-KEY> with your actual Cloudglue API key.

5. If you are building locally for integration with Claude Desktop or other clients, you may also build the server and extension:

   • Clone the repository
   • Run: npm install
   • Run: npm build

6. Configure your MCP client to point at the running MCP server as shown in the README examples for your client.

Notes and tips:

• The API key is required to access Cloudglue data; ensure you keep it secure and do not commit it to source control.
• The server exposes a suite of tools for discovery (list_collections, list_videos) and per-video analysis (describe_video, extract_video_entities). Familiarize yourself with pagination and time-range parameters described in the README to fetch large results incrementally.
• When configuring MCP clients, you may specify the path to the built server (for example, in the client JSON) if running locally. The README provides example JSON structures for both the npx-based startup and local builds.
• If you encounter rate limits or API quota issues, consider staggering requests or paging through results as described in the tool documentation.
• For Claude Desktop integration, follow the Desktop Extension steps in the README to install and authorize with your API key.
• Environment variables can be used to store sensitive keys; if you supply the API key via env vars, ensure your MCP client supports reading from env and that the value is kept secure.